EU executive weighs ditching US cloud and Anthropic

The European Commission is weighing replacing the US cloud provider and the Anthropic large language model that underpin its Job Matching Application, European outlet Euractiv reported.

Euractiv’s reporting says Commission staff have raised privacy and data‑sovereignty concerns over hosting candidate information and running the recruitment model on US infrastructure and a US‑based AI provider. Those internal objections have prompted officials to examine whether a so‑called ‘sovereign’ European cloud and European AI models would be more appropriate for the pilot application.

Commission officials developed the Job Matching Application as part of a broader drive to help public employment services and member states match jobseekers to vacancies, using automated ranking and suggestion tools. The prospect of swapping underlying infrastructure emerged after a review of the project’s data flows and third‑party dependencies flagged potential exposure of personal data to jurisdictions outside the EU, Euractiv reports.

The move would be a rare, concrete instance of an EU institution making procurement choices driven by data‑sovereignty priorities and concern about algorithmic risks. European policymakers have been explicit about reducing reliance on non‑EU cloud giants for sensitive public‑sector workloads; the Commission’s internal deliberations underscore how those policy aims are filtering into operational decisions on tools that process worker and candidate data.

For HR leaders, the case highlights two linked risks in AI‑driven recruitment systems: where data is stored and processed, and which models are used to score or rank candidates. Data protection experts say cross‑border processing can complicate compliance with the GDPR, while algorithmic decision‑making in hiring has repeatedly drawn scrutiny for disparate impacts and opaque score calculations.

The potential switch also fits a market trend: a number of European governments and institutions have signalled a preference for regionally controlled cloud infrastructure and for AI models trained and governed under EU legal frameworks. That political and regulatory pressure has begun to shape procurement and product road maps for talent‑technology vendors, which must balance performance, cost and legal risk when choosing partners.

Euractiv’s piece does not say whether the Commission has settled on alternative providers, which European cloud vendors are being considered, or how quickly any migration would take place. The reporting also leaves open key compliance questions that matter to HR teams and public‑sector buyers: whether any replacement models would be subject to independent bias audits, how data portability and deletion requests from candidates would be handled, and what contractual changes would govern data transfers.

Cost and vendor‑lock‑in are additional unknowns. Moving core services from a well‑established US cloud and a commercial LLM to ‘sovereign’ alternatives can entail significant engineering work and higher operating costs, and the Commission has not disclosed estimates or a procurement timetable.

If the Commission does proceed, the decision could influence private employers and HR technology suppliers that are weighing similar privacy and legal risks when deploying recruitment algorithms. A high‑profile institutional swap would signal that procurement choices can be driven as much by legal and political considerations as by model performance — an important signal for HR teams deciding whether to host candidate data domestically or in third‑party clouds, and for vendors designing redress, transparency and auditing features into their products.